Modelling Train Accidents East Java

IOSR Journal of Mechanical and Civil Engineering (IOSR-JMCE)
e-ISSN: 2278-1684,p-ISSN: 2320-334X, Volume 7, Issue 2 (May. - Jun. 2013), PP 07-16
www.iosrjournals.org
www.iosrjournals.org 7 | Page
Modelling The Train Accidents At Railroad Crossings
In East Java
Nyono1
, Harnen Sulistio2
, Achmad Wicaksono3
, Ludfi Djakfar3
1
Student at Doctoral Program, Departement of Civil Engineering, Faculty of Engineering, University of
Brawijaya,
2
Professor at Doctoral Program, Departement of Civil Engineering, Faculty of Engineering, University of
Brawijaya,3
Lecture at Doctoral Program, Departement of Civil Engineering, Faculty of Engineering,
University of Brawijaya (Indonesia)
Abstract: The growing train movement and people activities around the railroad will increase the frequency of
traffic in railroad crossing. This potentially results in the increase in traffic accidents. The prediction of the
number of such accidents is influenced by some factors dealing with variables on sensory psychological
behaviors and the perception of the drivers passing the crossings. Observations were made at 33 points railroad
crossing with not guardrail in East Surabaya DAOP VIII. The responsive variables are determined by the
explaining variables namely the number of train accidents in railroad crossing. The explaining variables are
those determining the value of responsive variables, consisting of three factors namely train engineering
features, road engineering features and environment. The last Poisson regression model possesses four
determining variables significant with the number of accidents, that is the train speed, the distance of signs and
the railroad crossing, flashing lamps and the average daily traffic. The train speed seems to be a primary factor
contributing to the high level of accidents. The results of sensivity analysis show that if the train speed
increases of 50%, the number of accidents will increase 40%. Facilities that should be quickly provided are
among others: provision and installation of flashing lamps and EarlyWarning System (EWS).
Key word: railroad crossing, train engineering features, road engineering features, environment factor,
Poisson regression
I. Introduction
The System of railway affairs in East Java has been established since the Dutch collonialism era. The
lines of train in East Java consist of North Line (Surabaya Pasar Turi – Semarang – Jakarta), Central Line
(Surabaya Gubeng – Yogyakarta – Jakarta), South Ring Line (Surabaya Gubeng - Malang – Blitar – Kertosono
– Surabaya) and East Line (Surabaya Gubeng – Jember – Banyuwangi). This province also possesses a
transportation system of commuter trains with a route of Surabaya – Sidoarjo – Porong, Surabaya – Lamongan
– Babat, Surabaya – Mojokerto, and Malang – Kepanjen. The train movement in each operation area (herein
called DAOP), each DAOP VII Madiun, Daop VIII Surabaya and DAOP IX Jember is high enough, which
result in a complicated problem and one of its negative effects in the increasing number of train movement in
east Java is accidents. In East Java, there are 1441 railroad crossings consisting of 1103 crossings without
guards, 338 with guards and also gate and 96 illegal crossings (PT.Kereta Api Indonesia, 2010), and the
potency to open or to add new railroad crossings is very great, especially the opening of illegal crossings due to
the growth of hinterland in either the right or left side railroad because of the growing land use in each railroad
areas. The growing train movement and people activities around the railroad will increase the frequency of
traffic in railroad crossing. This potentially results in the increase in traffic accidents.
Train accidents in railroad crossing often happen in line with the time development. The prediction of
the number of such accidents is influenced by some factors dealing with variables on sensory psychological
behaviors and the perception of the drivers passing the crossings (Raslear, 1996); categories of warning
equiptments, volume of road traffics, volume of train traffic, visibility of the condition in the crossings
(Gitelman and Hakkert 1996); types of warning equiptments, crossing geometric, railroad geometric, volume
of rtaffic (Saccomanno, Liping Fu and Moreno 2001); the number of the passing train, active equiptment, road
safety, rescuing operation, warning sign of flickering lamps, (Mok and Savage 2003); width of crossing
geometric, traffic control equipment, flickering lamp time, speed in heaping land, size of crossing, warning
signs, stop sign, number of railroad, number of tract, diameter of road separator, audit of safety, AADT,
warnign equiptment, control management, barrier control, status of class of road, types of area aroung the
crossing (business, residence, agriculture, etc)(Kang Lee and Ren Hu 2007); number of train identification,
levels of service, types of vehicles involved, number of damage of vehicles, number of the people injure or die
(Collister and Flaum 2007); factors of engineering in the crossings, of human beings, of environment (Zaharah
Ishak 2007); traffic separator, behavior or drivers’ responses factors to the equipments in railroad crossings

Modelling The Train Accidents At Railroad Crossings In East Java
(Ko, Washbum, Courage dan Dowell 2007); volume of traffic and trains per hour, speed of vehicles
approaching the crossings, percentage of heavy vehicles, levels of service (LOS), speed of the train approacing
the crossings (Zaharah Ishak,Yue and Somenahalli 2010); features of trains, roads, railroad crossings and of
traffic (RenHu,ShangLi and KangLee 2011). The above variables really influence the prediction of accidents
in railroad crossings, therefore some of the variables that influence one another may be simulated intro a model
of prediction of accidents in railroad crossings.
Various models of prediction of train accidents in railroad crossing have been developed. Federal
Railroad Adminstration (FRA) of America has studied accidents in railroad crossings by accomodating variables
among others multiplication of the average daily traffic factor in roads and traffic of the trains that passed, the
number of the passing trains per day, the speed of the trains, the number of tracts, the number of lanes in roads
and types of road hardening prove to influence the number of accidents in railroad crossings. Empirical results
show that the Poisson regresion is appropriate for estimaing the possibility of accidents; and the negative
binominal regression is good for predicting accident risks and effects (Kang Lee dan Ren Hu 2007). The model
was developed using a Petri Nets approach by taking into account components of basic concepts of safety,
infrastructure engineering techniques, levels of surrounding enviroment and all factors in human beings
(Zaharah Ishak,Yue dan Somenahalli 2010). The zero Possion regression model has also been developed to
delienate the relationship between the number of zero death or injury, and additional data and explaining
variables were collected in 592 locations of Railroad Grade Crossing (RGC) in Taiwan (RenHu,ShangLi dan
KangLee 2011).
Up to now, no research has been made to make a model of prediction of train accidents in railroad
crossing with no gate by accomodating and combining and developing all explaining variables that have once
been studied by previous researchers with different analyses. The resulted model would be built to predict train
accidents in legal railroad crossings without gate when a train is moving in a single track which generally
happens in developing countries with the minimal level of the society on the safety of trains – this makes them
easy to open ilegal railroad crossings. The results of this present study was expected to be useful for making any
action programs to reduce the number of train accidents.
II. Theories And Method
The variables of this present research consist of explaining and responsive variables. The responsive
variables are determined by the explaining variables namely the number of train accidents in railroad crossing.
The explaining variables are those determining the value of responsive variables, consisting of three factors
namely train engineering features, road engineering features and environment. The train engineering features
factor contains variables of the width of crosing, number of tract, speed of train, volume of passing train, free
vision of the engineer of locomotive, guardril in the crossing , the existence of flashing lamp and siren. The
road engineering features factors consist of agricultural areas, business, residence, industrial and road lights. In
the modelling of the number of accidents, a Poisson regresion analysis calculated using a statistical software
GenStat Discovery Edition 3 is employed. The stages of the data analyses are as follows:
1. Testing the distribution in the response variable (Y) using the Kolmogorov-Smirnov test of the data on the
number of accidents. In this test, the data of the number of accidents are expected to follow the that of
Poisson.
2. Establishing the model of the Poisson regression with a general model of 𝜇 = exp⁡(𝛽0 + 𝛽1 𝑋1 + … + 𝛽𝑘 𝑋𝑘 ),
making the following steps:
a. Modelling the Poisson regression analysis in each explaining variables (independent variables).
Estimating parameters for each combination of the Poisson regession model, with a general model
𝜇 𝑥𝑖, 𝛽 = exp(𝛽0 + 𝛽1 𝑋1𝑖 + ⋯+ 𝛽𝑘 𝑋𝑘𝑖 ), dengan i = 1, 2, 3, 4, 5.
b. Testing the parameters in each combination using Chi-Square, in GenStat Software this value are
transformed to F dan t statistics.
c. Testing the model for each combinations simultanesously with deviance criteria.
d. Determining the best model with the smalles deviance from each combination of variables.
e. Making an interpretation of the best model.
III. Research Results
The data distribution following the Poisson distribuition shows specific characteristics among others it
is discrite and limited in time or certain areas. The accident posisbility is very small, meaning the any vehicles
passing railroad crossings have a very small possibility to get accidents. The average number of accidents is 1,
45, meaning that in the last three years the number of accidents is about 1 -2 times in one point. From the
results of the Kolmogorov Smirnov test, the obtained value of the Kolmogorov-Smirnov Z is 1.341 with
asymp.Sig. (2-tailed) or p-value of 0,055. It may be concluded that the data on the number of train accidents
followed the Poisson distribution. The modelling showing the relationship among the numbers of train

accidents in railroad crossings would be made by using the Poisson regression analysis. The analysis consists of
three stages, first making a model for each explaining variables, second, modelling a combination of variables
that prove to have a significant influence from the results in the first stage, and three, selecting the determining
variables in the second stage which are really significant as a whole.
In a model with single determining factor, an analysis for significant 15 explaining variables from the
results of a descriptive analysis on the number of accidents is made. In the analysis, an emphasis is given on the
results of the regression coefficient test. If the result is significant (probability value < 0,05), this variable will
be included in the establishment of the simultaneous model. The analysis of each variable is presented in Table
1 :
Tabel 1. A Poisson Regression Analysis of the Influence of the Train Speed
Variable estimate s.e. t(*) p-value. Test results
Train speed 0.0311 0.0115 2.69 0.007 Significant
Train Volume 0.0558 0.0255 2.19 0.029 Significant
Signs 0.223 0.436 0.51 0.609 Not Significant
Distance of signs -0.0248 0.0126 -1.96 0.049 Significant
Free view -0.00118 0.000592 -2.00 0.046 Significant
Guardrail 0.182 0.298 0.61 0.541 Not Significant
Flashing Lamp -0.598 0.29 -2.06 0.039 Significant
Road width 0.362 0.182 2.00 0.046 Significant
Number of lane 0.171 0.472 0.36 0.718 Not Significant
the average daily
traffic 0.001436 0.000577 2.49 0.013 Significant
Road flatness 0.223 0.436 0.51 0.609 Not Significant
Types of construction 0.633 0.306 2.07 0.039 Significant
Road marks 0.27 0.295 0.92 0.360 Not Significant
Environment 0.598 0.29 2.06 0.039 Significant
Lighting 0.266 0.333 0.80 0.425 Not Significant
From the result of modelling with a single determining factors of 15 explaining variables, there are 9 (nine)
variables with significant influence, mean while the rest (6 variables) do not give any significant influence. Then
a simultaneous model involving 9 the (nine) significant variables are analysed. The results of the Poisson
regression analysis enclosing the 9 determining factors filtered in the first phase are shown in Table 2.
Table 2. An Analysis of the Poisson Regression of the 9 Chosen Variables
Variable estimate s.e. t(*) p-value Test results
Constant -0.79200 0.52400 -1.510 0.145 -
Train speed 0.01117 0.00647 1.730 0.098 Not Significant
Train Volume 0.00969 0.00882 1.100 0.283 Not Significant
Free view -0.00042 0.00023 -1.850 0.078 Not Significant
Flashing lamps -0.14700 0.13300 -1.110 0.280 Not Significant
Width of road 0.11570 0.08140 1.420 0.168 Not Significant
the average daily traffic 0.00065 0.00037 1.760 0.092 Not Significant
Types of construction 0.04500 0.14300 0.310 0.757 Not Significant
Environment 0.03600 0.13900 0.260 0.800 Not Significant
The last stage is intended to establish a regression model significant to the level of accidents simultaneously or
partially. The selection of such a model is made by excluding variables one by one that partially does not
influence the level of the accidents. From the results of the selection, there are four variables with significance
of 0.05 namely: the train speed (X3), the distance of signs and the railroad crossing (X7), flashing lamps (X10)
and the average daily traffic (X14), that significantly influence the level of accidents.

0
1
2
3
4
5
0 3 6 9 12 15 18 21 24 27 30 33
AmountOfAccident
Location
Actual Prediction
Table 3. A Final Model of the Results of the Poisson Regression Analysis
Variable estimate s.e. t(*) t pr. Test results
Constant -0.5910 0.3670 -1.610 0.118 -
Train speed 0.0130 0.0044 2.960 0.006 Significant
Flashing lamp -0.2575 0.0953 -2.700 0.012 Significant
the average daily
traffic 0.0011 0.0002 5.680 0.001 Significant
The results of the analysis of the four chosen variables show the p-values of less than 0.005. therefore, the best
model has been obtained. The following is the results of the analysis using the Poisson regression equation:
Y = exp(-0,591 +0,01302 Train speed – 0,01253 Distance Of Signs – 0,2575 Flashing lamp +0,001122 Average
daily traffic))
The validation of the model will measure the level of appropriatennes of the model with the results of real
observation. The results of such validation may be considered through the results of analysis of deviation
between the estimated value and the real value, the correlation value and the deviation test between the results of
prediction and real values. From the results of validation and the prediction value and the number of real
accidents, it seems that they seem not too different. The following is presented the results of the deviation
analysis from the last model and the picture of accidents prediction and the number of real accidents in each
point.
Picture 1. Values of Prediction of the Number of Accidents Based on the Poisson Regression Model
Picture 1 shows that the results of the prediction of the number of accidents reach the actual value. If the
prediction value of the number of accident is rounded, there are 27 points (81.8%) posessing the same value
between the prediction and the actual ones and 6 other points show differences. The different points are among
others the sample points no. 11, 18, 20, 22, 23 and 29. The validation of other models was made by calculating
the results of the deviation test of the number of accidents between the prediction and actual values. The test
was made using the paired- t-test. As in the Table 5.34, the average difference between the prediction and
actual values are -0.0303 with the p-value of 0.662 (higher than 0.05). It can be concluded that there is no
significant difference between the prediction value (from the Poisson regression model) and the actual value..
The last Poisson regression model possesses four determining variables significant with the number of
accidents. On the basis of the obtained model, decreasing the number of accidents can be reached with the
following ways:
 Reducing the train speed when passing railroad crossings.
 Putting in signs in a greater distance before the point of crossing
 Maintaining and keeping flashsing lamps to make them function well
 Giving special attention at morning and afternoon peak hour daily traffic with guard in the railroad crossing
with not guardrail.
In this part, an analysis of the level of sensitivity of a determining variable is made with the assumption that the
condition of other variables are unchanged. The analysis of sensitivity will be graphically shown by splitting the

0.0
0.5
1.0
1.5
2.0
2.5
0 200 400 600 800
AmountOfAccident
Average Daily Traffic (pcu)
No Flashing Lamp Flashing Lamp
0.00
0.50
1.00
1.50
2.00
2.50
3.00
40 60 80 100 120 140
AmountOfAccident
Train Speed (km/hour)
0.00
0.50
1.00
1.50
2.00
0.0 10.0 20.0 30.0 40.0
AmountOfAccident
DistanceOf Signs (m)
data into two conditions: availability and unavailability of flashing lamps. The trains speed is around 65 – 90
kms/h, the distance between the sign and the crossing is about 3 – 40 meters, meanwhile the value of the
average daily traffic is from 33.8 – 919.6 smp. The level of accidents is too low if in the crossing flashing lamp
is available, the train moves in low speed, yje distance between the sign and the crossing is far enough from
the crossing point and the daily traffic is not intense.
Picture 2. Amount Accident Prediction at few of
Velocity
Picture 3. Amount Accident Prediction at few
distance of Sign.
Picture 4. Amount Accident Prediction at few of the
average daily traffic

From the results of the sensitivity analysis on the basis of the changes in the train speed and in the number of
accidents, it is known that if the increase in the train speed reaches 50%, the number of accidents is predicted to
increase around 40%. If the increase is 100%, the number of accidents is predicted to be about 90%. And in the
results of the analysis of sensivity of the changes in the distance of the signs and the crossing and in the
number of accidents, it is shown that the decrease in the distance of the sign and the crossing point to 50% is
predicted to be able to increase the number of accidents up to 20%. If the distance is reduced up to 75%, the
number of accidents is predicted to increase about 35%. From the results of the analysis of sensitivity to the
changes of the average daily traffic up to 100%, it is predicted to increase the number of accidents of about 8%.
If the value of the average daily traffic increases up to 200%, the number of accidents is predicted to increase of
17%.
IV. Discussion
From the results of modelling with the Poisson regression analysis, there are four independent
variables found to be significant in the model. The variables are the train speed, flashing lamps, the distance
between the signs and the crossing, and the average number of daily traffic. The application of the model of the
Poisson regression has a high validation, which is in line with a research Chi-Lee and Ren-Hu (2007) made that
the Poisson regression is good for predicting the possibility of accidents; and the negative binomial regression
is good for predicting the risks and effects of accidents.
Average daily vehicles passing railroad crossing has a correlation with the number of traffic accidents.
The higher the average vehicle crossing the railroad in a year, the higher the number of accidents in raildroad
crossings. This also applies to the train speed and the possibility of accidents. Although human factors play a
big role in accidents, but this shows that the train, roads and environment features give a big contribution to
accidents in railroad crossings.
In the last model, it can be explained that the train features factor is important since from the four determining
variables in the number of accidents, three of which are train features such as train speed, flashing lamps, and
the distance between the signs and the railroad crossings. Road feature factors are representend by the average
daily number of traffics. While the environment factors condsisting of crossing areas (agriculture, housing and
industry) are included into the last model. In the process of modelling, from the results of analysis in the first
pat, it is evaluated that in the single determining model, the following variables give significant influence on the
number of accidents in railroad crossing, namely:
 Train speed
 Train Volume
 The distance between the signs and the railroad crossings
 Free view
 flashing Lamp
 Width of road
 the average daily traffic
 Types of construction
 Environment
But in the advanced model combining all single determining variables, there are merely four variables that prove
to have significant influence on the number of accidents. There are five variables considered as variables with
strong potentials as the causes of accidents namely:
 Train volume
 Free view
 Width of road
 The average daily traffic
 Types of construction
 Environment
A train feature important to hamper accidents is flashing lamp. This tool contained in the siren, both of which
function together. In the whole data, the number of railroad crossings with flashing lamps is 14 points (42,4%)
while the rest, 19 points (57.6%) without flashing lamps. The importance of the control equiptment is in line
with a research Coleman (1997) made. The control equipment of trafic provides passive-static warning,
guidance and in some cases, obligatory action for drivers. Trafic control equipments are assets that give warning
that the train is approaching. They are activated by a train in the circuit of tract/rail detection. This active control
equiptments are prodived with the same signs of crossings to give a passive control.
The railroad crossings provided with flashing lamps are 14 points, 13 of which (92.9%) became places
for accidents once in three years, meanwhile in one other point, 3 times accidents happened. Different from
points of observation without flashing lamps amounting 19 points, there are 2 points (10.5%) with a high level

of accidents namely 4 incidents during 3 years. This automatic instalation will help reduce the number of
accidents. The research results support the research made by Mok and Savage (2003). From the analysis, it can
be concluded that the instalation in the Guardrail or flashing lamps contributed about the fifth of reducing the
number of accidents. The development from a campaign ―safey operation‖ intended to inform the public about
proper attitudes in railroad crossings has long been made. In the 1970s and early 1980s, an instalation of ―ditch
lamp‖ in the locomotive has been known.
At present there is a wireless technology of early warning tool in railroad crossings. Due to the development of
transportation technology and provision of supporting facilities and infrastructures, a system of transportation
arrangement will be needed to improve pleasure and safety of the users of transportation facilities, especially in
railroad crossings which are not provided with guardrails. The tool of early warning in railroad crossings has
made use of a wireless system, so no cable is needed. This tool may turn on after the sensor works when the
traiin will pass the point in one km before it and it will transmit a sign to the warning tool to turn on. This
system also uses electric power from solar cells so that it will not depend on the electric supply from the state
electricity enterprise. Therefore it may be used in most railroad crossings located in a tract with no electric
network.
The train speed seems to be a primary factor contributing to the high level of accidents. The results of
sensivity analysis show that if the train speed increases of 50%, the number of accidents will increase 40%. The
level of sensitivity of this train speed is far superior than the distance between the signs and the crossings
namely 20% and the average daily traffic which is merely of 6%. This result is in line with that of Coofster and
Pflaum (2007) stating that the explaining variables significantly influencing the possibility of accidents are the
train speed, the number of trains passing the crossings each day, the percentage of heavy vehicles (trucks), the
number of vehicle traffic (number of lanes), signs on the roads, flashing lamps, railcrossing angle, the surface
of road and railroad crossings, trade, housing and industrial areas. The problem of train speed in Indonesia is
dilemmatic. Reducing the train speed will result in the addition of movement time, whereas without reducing
the speed, the train often comes late in its destination. It seems in contradiction with the development of railway
affairs in other countries, where the train speed has always been added. In Indonesia, it it PT KAI that operates
the trains. While the facilities and infrastructure are handled by the government.. Reducing train speed should be
made due to bad condition of the track, and this condition happens because fund allocation from the
government to maintain the track is very small compared with what actually needs. As a result, from year to
year, the condition of the track will be worse. As an operator, reducing the speed is an appropriate choice since
if an accident happens, it is the operator (PT KAI) that will be responsible for it. Up to now, the maintenance of
the tracks are still held by PT KAI as an operator.
The results of modelling using the Poisson regression will be used to predict the point at which a
railroad crossing should be paid attention. A ―blackspot‖ status for a railroad crossing with high level of
accidents will be able to help reduce accidents. A blackspot is a crossing with high risk of collision. It is
suggested that one of the way is to allocate fund for all fields of problems. A random incident of collision is
very various in space and time. A high risk in a certain crossing in a year does not always show high risk in the
next year. A risk of collision needed to express any risk may be anticipated in a certain period. This estimation
may be obtained using a model to rpedict a frequency of collission and therefore it is accurate and reliable. The
identification of a blackspot merely based on the number of collission will not give any complete picture of the
risk in each crossing. The risk of collission consists of two components: frequency and consequence (level o
severity). Ignoring such a consequency may result in less intervention in any railroad crossing with the severe
level of collission and riska -based model is needed to identify any spots where collissions often happen.
The results of prediction of the number of accidents in each spot may be used to attribute certain characteristics
ot the spot. Another indicator to choose the best criteria is comparing the number of expected accidents and that
of observed accodents (Rakhmat et al., 2012). The results of the comparison may be in the form of
a. Location which is predictied to be dangerous is actually harmful (correct positive)
b. Location which is predictied not to be dangerous is actually not harmful (correct negative)
c. Location which is predictied to be dangerous is actually not harmful (false positive)
d. Location which is predicted not to be dangerous is actually harmful (false negative)
In this case, if the observed number of accidents is higher than the expected one, it can be categorised into
correct positive (CP). If the observed number of accidents is lower than the expected one, it is categorized as
false positive (FP). In Table 5.40, it is shown that the criteria excessive number of accidents using the
prediction model resilts a number of segment classified as the biggest correct positive (CP) as compared with
the other three criteria, namely 7 segments (from 10 mist dangerous segments) and 14 segments (from twenty
most dangerous segments).

Table 4. A Comparison of Actual and Predicted Values from the Poisson Model
Spot Actual Predicted Information Spot Actual Predicted Information
1 1 1.45 False Positive 18 4 3.13 Correct Positive
2 2 2.31 False Positive 19 1 1.04 False Positive
3 2 1.51 Correct Positive 20 1 1.92 False Positive
6 1 0.86 Correct Positive 23 3 2.45 Correct Positive
12 1 0.94 Correct Positive 29 4 3.29 Correct Positive
17 1 0.90 Correct Positive
In Table 5.40, it is shown that there are 10 spots which are really dangerous namely spot 3 (Bojonegoro regency
; 140+135, SRJ-BWO), spot 6 (Lamongan regency; 162+681, BBT-GEB), spot 12 ( Lamongan regency;
179+735, SLR-LMG), spot 17 ( Gresik regency; 199+790, LMG-DD), spot 18 (Surabaya city; 222+603, KDA-
TES), spot 23 (Sidoarjo regency; 26+121, SPJ-BH), spot 28 ( Pasuruan regency; 43+629,, PR-BG), spot 29 (
Pasuruan regency, 44+610, PR-BG), spot 30 ( Malang regency; 29+128, SN-LW) and spot 32 ( Blitar
regency; 76+158, NB-SBP).
In this research fangerous segments are determined by comparing three criterias with the data of train accidents
from 2010-2012 and applying the resulted model to get the expectation of average number of accidents on the
referred population. The criteria to determine the dangerous segments that will be used are among others:
a. The excess of the number of accidents using the model of traffic prediction is made by determining the
difference of number of accidents from the prediction model and the results of the observation.
b. The level of accident is mad by comparing the real number of accidents and the daily crossing of a
segment,
c. The frequency of accidents is made by ordering the data on traffic accidents from the highes to the lowest.
d. The results of identification of all examined spots using the criteria of the excess number of accidents create
a rating of dangerous segments based on the difference between the observede number of accidents and the
expexted results of the prediction model. The result of discussion on the basis of the final model of the
Poisson regression analysis will results in some implications intended to reduce the level of accdients.
To avoid any collision between the train and general transportation in railroad crossings is made by applying a
technology to improve the reliability of the signals, either those in any crossing with/without guard. The
available and proper technology for the purpose is installing the AWS (Automatics early Warning System).
Since there are thousands of railroad crossings with no automatic gate, it is proper to apply the AOCL
(Automatic Open Crossing, Locally monitored) since it is cost effective. Besides the application of the
technologies, other efforts which should be simultaneously applied are as follows:
a. Completing traffic signs on the roads that will cross the railroad crossings.
b. Controlling any railroad crossing by closing or combining two or more crossing into one.
c. Reducing railroad crossing using flyover or underpass
An alarm system in railroad crossing is used by providing flashing lamps and sirens. In each railroad crossing
with or without gate, signs and alam/sirens should be provided, since the most effective sense is ears (earing),
and ears can respond information without being able to be caught by sense of sight, especially in any crossing
surrounded by high buildings. Psychologically, if alarm (siren) is heard, there is tendency for one sense to be
more alert than others. For example, for the sense of sight (eyes) although they have seen any written warning,
but there is a tendency that the influence of impatience is still higher. Alarm or siren should be placed in each

crossing especially those with no gate. It is better any censor or switch alarm/indicator lamp are put 500 mt
from the crossing, so that drivers may quickly know the position of the train to take any step to avoid accidents.
V. Conclusion
From the discussion above, some conclusions can be made. The train accidents happening in railroad
crossings without guardrails in the operational area of DAOP VIII Surabaya for the last 3 (three) years, from
2010 to 2012, are 149 incidents with the following characteristics: hit by persons, by motor cycle (R2) and by
personal vehicles or truck (R4) with the death of 30 persons, injuries, 107 persons and no victims of 12 persons.
From the results of modelling the Poisson regresion, there are four determining factors of accidents namely
train speed, the distance of the signs and the railroad crossing, flashing lamps and the average daily traffic. The
train speed possesses the highest sensitivity to the number of accidents.
From the results of modelling to the number of accidents, some reccomendations are offered. Installing
flashing lamps in each railroad crossing with not guardrail proves to contribute to the decrease in the number of
accidents. So it is recommended that in each railroad crossing without guardrail be put in flashing lamps. It
should optimize the participation of the people living around railroad crossings to maintain the warning signs or
other safety facilities in railroad crossing with no gate. Any activity of socialization to people living around
the crossings should be made in roder to improve their participation in keeping the security and safety in the
crossings.
Technical guidance to the people should also be given to improve the participation of the people living
around railroad crossings with not guardrail. The operator should make a stronger coordinating with the
concerned institutions to improve safety and security in the railroad crossing with no guardrails because of
limited budget from the operator to providing the safety facilities. Facilities that should be quickly provided are
among others: provision and installation of flashing lamps and Early Warning System (EWS), signs in certain
distances (not too near or too far), signs of speed limit and of signal 35 for each railroad crossing with no gate
and inspection and control are made in cooperation with concerned institutions to close any illegal railroad with
no gate.
References
[1]. Arikunto, S. 2010, “Prosedur Penelitian Suatu Pendekatan Praktek” Gramedia Pustaka Utama Jakarta.
[2]. Coleman, Eck and Russell 1997, ―Railroad-Highway Grade Crossings A Look Forward‖ Committee on Railroad-Highway Grade
Crossings;
[3]. Collister and Pflaum 2007, ‖A model to predict the probability of highway rail crossing accidents‖, Overland Kansas USA;
[4]. Direktorat Jenderal Perhubungan Darat (2005), ―Pedoman Teknis Perlintasan Sebidang antara Jalan dengan Jalur Kereta Api‖
Jakarta.
[5]. Farr 1987, ―Summary Of The DOT Rail—Highway Crossing‖ resource allocation procedure-revisited. Report no. DOT/FRA/OS-
87/05, Office of Safety Federal Railroad Administration.
[6]. Federal Railroad Administration (2001), ― Office of Policy and Program Development Grade ― Dec 2000 user manual, version 2.0.
[7]. Federal Railroad Administration (2002), Office of Policy and Program Development. Grade Dec 2000, Reference manual version
2.0, January 2002.
[8]. Federal Highway Administration (2000), ― Statistical Models of At – Grade Intersection Accidents – Addendum― March 2000
Publication No. FHWA - RD - 99 -094.
[9]. Gitelman And Hakkert 1997, ― The Evaluation Of Road-Rail Crossing Safety With Limited Accident Statistics‖ Transportation
Research Institute, Technion-Israel Institute Of Technology, Haifa 32000, Israel.
[10]. Hogg, and craig. 1978. Introduction to Mathematical Statistic. Fourth Edition. USA: Macmillan Publishing co, Inc.
[11]. Hauer 1985, ―On The Estimation Of The Expected Number Of Accidents‖ Transport Safety Studies Group, Department Of
Civil Engineering, University Of Toronto, Toronto, Ont., M5s La4, Canada.
[12]. Hensher and Button 2000, ―Hand Book Of Transport Modelling‖ , Netherlands;
[13]. Hanseon Cho and Rilett 2007, ―Improved Transition Preemption Strategy for Signalized Intersections near At-Grade Railway Grade
Crossing‖ DOI: 10.1061/ (ASCE)0733-947X (2007)133:8 (443);
[14]. Hair,Black, Babin dan Anderson 2010, Multivariate Data Analysis, Seventh Edition copy right 2010 by Pearsion Prentice Hall.
[15]. Hermanto Dwiatmoko (2011) ―Program Strategis Peningkatan Keselamatan Perkeretaapian”, Direktorat Jenderal Perkeretaapian
Kementrian Perhubungan Jakarta;
[16]. Kleinbaum, R.H.1988. Applied Regression Analysis and Other Multivariable Method. Boston: PWS-KENT Publishing Company.
[17]. Ko, Washburn, Courage and Dowell 2007, ―Evaluation of Flexible Trafﬁc Separators at Highway–Railroad Grade Crossings‖ DOI:
10.1061/(ASCE)0733-947X (2007);
[18]. Lee, Nam, and Park 2005, ‖Analyzing The Relationship Between Grade Crossing Elements And Accidents‖,’Journal of the Eastern
Asia Society for Transportation Studies, Vol. 6, pp. 3658 - 3668, 2005
[19]. Lee and Ren Hu 2007, ―Accident Risk At A Railway Level Crossing‖; Journal of The Eastern Asia Society for Transportation
Studies;
[20]. McCullagh, P, and J.A. Nelder. 1989. Generalized Linier Model. Second Edition. London: Champman and Hall.
[21]. Myers, R.H. 1990. Classical and Modern Regression with Applications. Edisi Keempat. USA: PWS-KENT Publishing Company.
[22]. Mok and Savage 2007, ―Why has Safety Improved at Rail – Highway Grade Crossing‖;Risk Analysis;
[23]. Nawari 2010, “Analisis Regresi” Alex Komputindo Kompas Gramedia Jakarta.
[24]. Peterman 2007, ―Selected Issue in Proposed Reauthorization Legislation‖;Federal Railroad Administration Programs;
[25]. PT. Kereta Api Indonesia 2010, Data Base Perkeretaapian, Surabaya.
[26]. Raslear1996, ―Driver Behavior At Rail-Highway Grade Crossings: A Signal Detection Theory Analysis‖; Washington, DC:U.S.
Department of Transportation.

[27]. Ren Hu, Shang Li and Kang Lee 2011, ―Assessing Casualty Risk of Railroad-Grade Crossing Crashes Using Zero-Inflated Poisson
Models‖ DOI: 10.106 1/(ASCE) TE.1943-543 6.0000243.© 2011 American Society of Civil Engineers ;
[28]. Rangkuti, F. (2004) ―Analisis SWOT Teknik Membedah Kasus Bisnis‖ Gramedia Pustaka Utama Jakarta.
[29]. Saccomanno, Liping Fu, and Moreno 2001, ‖Risk-Based Model for Identifying Highway – Rail Grade Crossing Blackspots‖
Transportation Research Record,’ Journal of the Transportation Research Board, No. 1862, TRB, National Research Council,
Washington, D.C.
[30]. Sian Tey and Ferreira 2010, ―Driver compliance at railway level crossings‖ Faculty of Engineering, Architecture and Information
Technology, The University of Queensland, AND Brisbane St Lucia, QLD 4072, Australia;
[31]. Singarimbun, M & Effendi, S. (1995) ― Metode Penelitian Survey‖ Jakarta LP3ES.
[32]. Yayasan Bhakti Ganesha (2005), Kecelakaan Kereta Api di Indonesia (Permasalahan dan Alternatif Solusi); ITB Bandung.
[33]. Zaharah Ishak 2007, ― The Development Of Railway Level Crossing Safety Assessment Model‖ Conference of Australian Institutes
of Transport Research (CAITR);
[34]. Zaharah Ishak, Yue and Somenahalli 2010, ―Level Crossing Modelling Using Petri Nets Approach and Π–Tool‖ Asian Transport
Studies, Volume 1, Issue 2 (2010), 107-121 © 2010 ATS All rights reserved;

Modelling Train Accidents East Java

Recommended

Recommended

More Related Content

What's hot

What's hot (19)

Viewers also liked

Viewers also liked (20)

Similar to Modelling Train Accidents East Java

Similar to Modelling Train Accidents East Java (20)

More from IOSR Journals

More from IOSR Journals (20)

Recently uploaded

Recently uploaded (20)

Modelling Train Accidents East Java